Conventionally, GPS/INS integrated navigation systems that calculate estimated positions, estimated velocities, estimated attitudes, estimated azimuths and the like of a moving body obtained by integrating sensors, such as a GPS, an inertial sensor, and a velocity sensor widely use a Kalman filter for error estimation. The estimation capability of the errors by the Kalman filter greatly influences model errors of a dynamic model or a measurement model, and particularly, the model errors differ greatly when the moving body is moving and is not moving. Further, in the stationary state, due to observable problems of the measurements, it may not be able to estimate for states of some errors, or the estimation accuracy may deteriorate. As a result, even though the moving body is stopped, a disadvantage may occur in which the positions, the velocities, the attitudes, and the azimuths vary.
In order to avoid the disadvantage, typical GPS/INS integrated navigation systems may be required to be constituted with a GPS receiver including multiple GPS antennas, and a highly-accurate three-axis accelerometer and a three-axis angular rate sensor; however, they are expensive (hundreds of thousands of yen to millions of yen). Particularly, for the land-based applications requiring a low-cost (tens of thousands of yen), configurations with a GPS receiver having a single GPS antenna, a low-cost single-axis inertial sensor, and a low-cost velocity sensor are desirable.
However, in the low-cost system configuration, degradation of the estimation capability due to at least the model errors, the observable problems, and particularly the weak observations in the stationary state cannot be avoided. Therefore, the disadvantage in which the position, the velocity, the attitude, and the azimuth vary will be remarkable even though the moving body is stopped.
As measures against the disadvantage which the low-cost system configuration has, a method adopting a measurement model equation is known, in which the velocity measurements, the angular rate measurements and the like are zero when the moving body is not moving (for example, Nonpatent Documents 1 and 2).
[Nonpatent Document 1] Anastasia Olegovna Salytcheva, “Medium Accuracy INS/GPS Integration in Various GPS Environments,” UCGE Reports Number 20200, Canada, UNIVERSITY OF CALGARY, 2004, pp 107
[Nonpatent Document 2] Yufeng Zhang and Yang Gao, “Observability Analysis of Initial Alignment and Its Accuracy Improvement,” ION GNSS 18th International Technical Meeting of the Satellite Division, 13-16 Sep. 2005, Long Beach, Calif., pp 1